KR100825960B1 - Method of assigning a landmark on a cephalometric radiograph - Google Patents

Abstract

The present invention comprises a first step of displaying a digital cephalometric radiograph on a monitor at the request of a user; A second step of displaying, in an image-enhanced form, a predetermined area including a point at which a cursor is located on the displayed digital head measurement radiograph; And a third step of recognizing the point designated by the user as the measurement point in the image-enhanced constant region. 2. The method of the present invention relates to a method for designating a measurement point on the head measurement radiograph.

Head measurement radiograph, tracing, image enhancement, measuring point, real time

Description

METHOD OF ASSIGNING A LANDMARK ON A CEPHALOMETRIC RADIOGRAPH}

1 is a diagram showing an example of a histogram pattern (a. Low clustering; dark image, b. High clustering; bright image, c. Small spread; low contrast, d. Wide spread; high contrast);

2 is a diagram illustrating an example of a histogram stretched image;

3 is a diagram illustrating an example of a histogram smoothed image;

4 is a diagram illustrating an example of an image gamma-converted (center: gamma = 0.5, right: gamma = 2.0);

5 is a view for explaining a method for performing a discrete line,

6 is a diagram illustrating an example of an image of a local region histogram smoothing process;

7 is a view showing an example of a screen configured for designating a measurement point according to the present invention;

8 is a diagram illustrating an example of an image to which an image enhancement technique according to the present invention is applied;

 9 shows evaluation results for Group A,

10 is a diagram showing an evaluation result for a system B group;

11 to 14 are graphs showing evaluation results.

The present invention relates to a method for designating a measurement point on a head-measuring radiograph, and more particularly, to a method for designating a measurement point in a local image-enhanced region by image-enhancing the digital head-measuring radiograph locally and in real time. .

A cephalometric radiograph is a standardized radiograph used for the development of diagnostic and treatment plans for orthodontic and orthognathic surgery, comparison before and after treatment, and research on growth and development. Head measurement radiographs are used for quantitative analysis and evaluation through the recognition of landmarks and distances, angles, and ratios of the landmarks, rather than reading the lesions through qualitative evaluation of the images. Therefore, accurate and correct location identification of the measurement point is the most basic and important part of the analysis of the head measurement radiograph. Recognition of these metrology points is centered on local anatomical structures around the metrology points, rather than referring to or reading the whole head radiograph.

Researches have been made to improve the accuracy and reproducibility of radiographic image reading using image enhancement techniques. Image reinforcement technique can be applied to digital tofu measurement radiographs, which can be used for recognition of measurement points, and some commercial analysis programs use it. The following are some of the video enhancement techniques.

Pixel Operations

This is a method of enhancing the image by adjusting the intensity of each pixel, that is, the radiation concentration, and does not consider the surrounding pixels or the entire image.

1.1. Histogram Stretching

1.1.1. Histogram

A histogram of a digital radiographic image is a tool used to grasp the characteristics of contrast values of an image. The histogram is a chart showing the distribution of intensity values of each pixel. The horizontal axis represents the intensity value, the vertical axis represents the frequency, and the histogram shows the frequency of each pixel having the intensity value.

A digital radiographic image can be thought of as a planar collection of pixel contrast values f (m, n) , where f () represents the contrast value of the pixel, and m and n represent the position of the horizontal and vertical coordinates of each pixel. do. In the case where the horizontal and vertical coordinates of the entire image are M and N, and the contrast ranges from 0 to P-1 and has gray levels of P steps, the histogram h (i) of each contrast value is It can be defined as

The histogram can be analyzed to determine the overall contrast distribution of the image (see FIG. 1). In addition, it is possible to obtain information on the applicability of image enhancement, in particular, whether good results can be obtained through the application of contrast enhancement. In the case of (a), (b), and (c) of FIG. 1, which are not uniformly distributed images as illustrated in FIG. 1d, the image may be converted into a high contrast image through histogram processing.

1.1.2. Histogram stretching

Histogram stretching is one of the useful tools to improve the quality of images with low contrast, and is also referred to as terms such as contrast stretching or intensity scaling. This is best applied when the distribution of the image histogram is skewed to a specific part rather than across the contrast, especially when the shape has a Gaussian distribution. In general, the entire contrast range of the image information stored in the digital image file is displayed on the screen according to the total gray scale that can be displayed on the computer screen, but histogram stretching is focused on a specific region rather than the entire contrast range. To observe.

When f ₁ and f ₂ represent minimum and maximum values of the region of interest, the histogram stretching can be expressed as follows.

Where f is the contrast value of the original image, e is the intermediate image value for conversion to represent only the region of interest, g is the value of the output image and f _max is the maximum contrast value of the computer screen.

In FIG. 2, an example of obtaining contrast-enhanced images using histogram stretching can be seen.

1.2. Histogram Equalization

Although histogram stretching is very effective in reinforcing image information of a specific area, it may be difficult to know in advance which area the useful information is in, or the histogram may be distributed in a form that is not appropriate for histogram stretching. A useful method in this case is histogram equalization.

Histogram smoothing is based on the theory that an image with a uniform histogram over an entire range of grays is ideal.

The ultimate goal of histogram smoothing is to create a histogram with a uniform distribution, not to flatten the histogram, but to redistribute the distribution of contrast values. Based on the histogram content defined in histogram stretching, each histogram should have (M × N) / P pixels for an ideally distributed histogram.

An easy way to redistribute the contrast is to use a normalized cumulative histogram. This can be defined by the following equation.

The normal cumulative histogram may be used as a mapping function for converting the contrast value of the original image into the contrast value of the enhanced image. The new histogram smoothed contrast value g (m, n) can be obtained by

3 shows an image with enhanced contrast through histogram smoothing. Histogram smoothing is effective when the image has fine details in dark areas. It is not desirable to perform histogram equalization on all images, since a good quality image can be changed by this. In FIG. 3, the overall contrast of the image was increased, but the soft tissue morphology around the nasion and the lower mandibular region became more difficult to observe.

1.3. Gamma correction

Gamma correction is a nonlinear transformation that is similar to the contrast control of a CRT monitor. This is often used to compensate for nonlinear features in screen or printer output and film printing. The general form of the gamma transform is:

g (m, n) = f (m, n) ^{1 / γ}

If the value of γ is 0, there is no change due to null conversion, and an exponential curve is formed in the range of 0 <γ <1 to darken the image. If γ> 1, a log-shaped curve is formed to brighten the image. The gamma of a typical CRT monitor is about 2.2 (range of 1.4 to 2.8). 4 shows the result of gamma conversion in which γ is set to 0.5 and 2. FIG. If you set it to 0.5, the whole image is dark, and if it is set to 2, it changes brightly and you can observe the nonlinear change.

2. Spatial Operations

Spatial operation is also called local operation, and it uses neighboring pixels as well as one pixel to create new pixel value through conversion. Most of the spatial operations are linear methods using kernel convolution, but some are non-linear.

2.1. Kernel convolution

This is not an individual image enhancement technique, but a mathematical formula underlying the linear spatial operation described later.

Discrete lines apply a square two-dimensional matrix called a convolution mask or kernel to each pixel and surrounding pixels in the image to obtain a new image (see FIG. 5). In this case, kernel is an odd-odd matrix, the center position should be equal to the position of the output pixel, and the sum of all coefficients in the kernel should be determined to be 1. The sum of the coefficients affects the overall brightness of the image. If it is 1, it has the same average brightness as the original image. If the sum of coefficients is 0, it is used for edge detection.

Let w (k, l) be each coefficient of the kernel, a matrix of size ( 2K + 1) × (2L + 1) , and point (k, l) = (0,0) is the kernel's coefficient. In the center, the calculation of the new contrast value g (m, n) by the discrete line using this kernel can be expressed by the following equation.

After matching the center of the kernel with the pixels m and n of the original image, the sum of the product of each coefficient of the kernel and the contrast value of the corresponding pixel is the contrast of the new image. The new image is obtained by repeatedly applying all pixels of the original image.

2.2. Edge enhancement

Edge enhancement is an important part of the image enhancement technique because the human visual system uses the contour as an important element for understanding the image content. Conventional psychophysical studies have reported that the human visual system tends to accept contour-enhanced photographs and radiographic images better than the original.

Horizontal and vertical contours can be detected and enhanced respectively. The image of which the contour is detected may be combined with the original image to obtain an image of which the contour is enhanced.

The horizontal outline can be detected using the following kernels.

In addition, by using the following kernels (kernels), it is possible to obtain an image with enhanced horizontal outline.

The vertical contour can be detected by applying the following kernels.

The image with enhanced vertical contour can be obtained by applying the following kernels.

The following kernel can detect the contour of the whole direction.

The contour reinforcement kernel in the overall direction using the above kernel is as follows. This kernel is also called an unsharp mask kernel because it blocks the low frequency region of the image and enhances the high frequency region to show the effect of sharpening. It is possible to strengthen the contours in all directions, but the degree of contour enhancement tends to be weak.

2. 3. Local Area Histogram Equalization

This technique applies the histogram smoothing described in the previous pixel operations. Instead, it applies to the partially divided local regions rather than the entire region of the image. This is a nonlinear processing technique, which improves the readability of the small details of the image.

Local area histogram smoothing can be applied to each pixel (m, n) of an image by applying the following equation.

Where h _LA (m, n) (i) is the histogram of the local region, H _LA (m, n) (j) is the cumulative histogram of the local region, and g (m, n) is the result of applying the local region histogram smoothing. .

This method requires a lot of mathematical calculations, which causes computational load on the computer, and the complexity of the operation increases in proportion to the square of the kernel size. Since the operation is applied only to the surrounding image of the local area, the change in the contrast value of the pixel is characteristically displayed, and the detail may be emphasized regardless of the size of the entire image. An important limitation is that the relationship between the input and output images is not linear and non-monotononic. This limitation means that even if the contrast value of the image is the same area (for example, the skeleton), it should be noted that it is converted into different contrast values according to the local area to be divided. 6 shows a local area histogram smoothing effect processed by arbitrarily dividing the area.

It is an object of the present invention to provide a method for designating a measurement point on a digital head measurement radiograph using a novel image enhancement technique.

It is also an object of the present invention to provide a method for designating a measurement point on a digital head measurement radiograph that can identify the measurement point within a locally image-enhanced region.

It is also an object of the present invention to provide a method for designating a measurement point on a digital head measurement radiograph capable of image enhancement locally and in real time.

Another object of the present invention is to provide a method for designating a measurement point on a digital head measurement radiograph having an enhanced image enhancement effect by image enhancement of a digital head measurement radiograph having a DICOM standard format.

To this end, the present invention comprises a first step of displaying a digital cephalometric radiograph on the monitor at the user's request; A second step of image-enhancing a predetermined area including a point where a cursor is positioned on the displayed digital head measurement radiograph; wherein the image enhancement is a real-time localized automatic histogram stretching technique, A second step made by one of a real-time localized automatic histogram equalization technique and an edge enhancement and real-time localized automatic histogram stretching technique; And a third step of recognizing the point designated by the user as the measurement point in the image-enhanced constant region. The method may include specifying the measurement point on the head measurement radiograph.

In another aspect, the present invention provides a method for specifying a measurement point on the head measurement radiograph, characterized in that the image-enhanced certain region is changed in accordance with the movement of the cursor.

In another aspect, the present invention provides a method for specifying a measurement point on the head measurement radiograph, characterized in that the image-enhanced constant area is centered on the cursor. If the image-enhanced area is too large, the system load will be large, and if the effect is reduced and if it is too small, it may not be helpful to identify the measurement point. × 2 cm).

In addition, the present invention provides real-time localized automatic histogram stretching, real-time localized automatic histogram equalization, and edge enhancement and real-time histogram stretching. A method of designating a measurement point on a head measurement radiograph, which is performed by one of time localized automatic histogram stretching) techniques, is provided.

delete

In addition, the present invention is characterized in that the digital head measurement radiograph before being displayed in the first step has a format conforming to the DICOM standard, and the image enhancement in the second step is performed on the digital head measurement radiograph following the DICOM standard. It is characterized by a method of designating a measurement point on the head measurement radiograph. Through such a configuration, it is possible to perform enhanced image enhancement than the image enhancement conventionally performed in 8-bit gradation.

Hereinafter, the present invention will be described in more detail with reference to the drawings.

1.1 Development of Real-time Local Image Enhancement Technique for Digital Head Radiograph

In order to recognize and designate measurement points by locally enhancing digital head radiographs in real time, the following image enhancement techniques were developed and applied independently or in combination.

1.1.1 Real time localized automatic histogram stretching

An image histogram of the region of interest is analyzed to obtain the lowest and highest contrast values in the region through an end in search. Stretch the histogram by limiting the region of interest to be set to 0 (black) for the lowest contrast in the region and 2 ⁿ -1 (white, where n represents the gradation of the image in bits) for the highest contrast in the region. Do this. The enhanced image is represented by implementing an 8-bit image of a PC screen through a linear transfer function.

If the lowest contrast value in the region of interest is f ₁ and the maximum contrast value is f ₂ , and the gray level of the original image is P , it can be expressed by the following equation.

Where f is a contrast value of the original image, e is a contrast value converted from P gray scale, and g is a contrast value of an image output on a PC screen having Q gray scale. P = 2 ¹² = 4096 for ^12- bit grayscale images, P = 2 ¹⁰ = 1024 for 10-bit grayscale images, and Q = 2 ⁸ = 256 because the PC used in the study represents 8-bit grayscales.

When applied in combination with other techniques, it is intended to be applied last to improve optimal contrast.

1.1.2. Real time localized automatic histogram equalization

The histogram smoothing of the image histogram is performed by limiting the region of interest centered on the measurement point. The enhanced image is represented by implementing an 8-bit image of a PC screen through a linear transfer function.

This can be expressed by the following equation for each pixel (m, n) in the ROI.

Where f (m, n) is the contrast value of (m, n) pixels, P is the gray level of the image, and Q is the gray level of the PC screen. h _LA (m, n) (i) is the histogram of the local area, H _LA (m, n) (j) is the cumulative histogram of the local area, and e (m, n) is the contrast in P gray after applying histogram smoothing. Value, and g (m, n) is the contrast value of the image output on the PC screen. The values of P and Q applied are as in 1.1.

1.1.3. Real time kernel convolution

Apply a 3x3 kernel to the region of interest. Kernel coefficients can be specified arbitrarily so that various discrete lines can be applied to the region of interest.

If w (k, l) is the coefficient of each kernel in a 3x3 matrix, and point (k, l) = (0,0) is the center of the kernel, the kernel is called The calculation by the discrete line used can be expressed by the following equation.

Where f (m, n) is the contrast value of (m, n) pixels, P is the gray level of the image, and Q is the gray level of the PC screen. e (m, n) is the contrast value in P grayscale after applying a discrete line, and g (m, n) is the brightness value of Q grayscale image output on the PC screen through a linear transfer function. The values of P and Q applied are as in 1.1.

When the discrete line is applied at the periphery of the ROI, the kernel is out of the ROI, but the image information outside the ROI is used when the kernel does not leave the range of the entire image. In a case where the entire image is out of range, a method of inserting a pixel having a contrast value of 0 is used.

1.1.4. Real time gamma correction

Using gamma values that can be arbitrarily specified, it is possible to perform gamma conversion limited to the region of interest.

e (m, n) is the contrast value of the original P gray level after applying the gamma transform, and g (m, n) is the contrast value of the Q gray level image output on the PC screen through a linear transfer function. to be. The values of P and Q applied are as in 1.1.

1.1.5. Inverse

The inverted image of the entire image and the ROI is expressed.

When the gray level on the PC screen is Q and the contrast value of the pixel (m, n) on the screen is g (m, n) , a new contrast value can be obtained by the following equation.

new g (m, n) = Q -old g (m, n)

By using the developed local image enhancement technique, a program for inputting measurement points in digital head measurement radiographs was made (see FIG. 7).

The developed program allows local image enhancement to be performed in a region of a specified size (1), and local image enhancement is processed in real time according to the movement of the cursor (2). Digital head measurement radiographs of the standard DICOM (Digital Imaging and Communication in Medicine) format were used, and image enhancement was performed on the images, and then displayed on the screen through a linear transfer function.

The operation of the program is performed by using a user interface (UI) provided at the right side of FIG. 7, when a user brings up a digital tofu-measuring radiograph, and then moves the cursor 2 to designate a measurement point. A local image enhanced region 1 is displayed, and the measurement point is found by finding the measurement point in this area 1 and inputting it using a keyboard or a mouse. In this case, the mode for displaying the local image enhanced region 1 may be configured to be performed at the same time as bringing up the digital head measurement radiograph on the screen, or may be operated by separate instructions. Meanwhile, a separate icon may be provided for conversion to another image enhancement technique, or may be configured to perform such conversion through input of a mouse or a keyboard.

1.2. Evaluation of the developed real-time local image enhancement technique

To evaluate the developed real-time local image enhancement technique, the following image enhancement techniques and image-enhanced images were used.

1.2.1. Original image without local image enhancement (Processing Technique 0).

Original images without local image enhancement were used and have the characteristics of a control group.

1.2.2. Image applying real-time local automatic histogram stretching (Processing Technique 1; see left figure in FIG. 8)

1.2.3. Image with real-time local automatic histogram smoothing (Processing Technique 2; see middle figure in FIG. 8)

1.2.4. Image with Edge Enhancement and Real-time Local Automatic Histogram Stretching (Processing Technique 3; see the right figure in FIG. 8)

It was used to investigate the effect of contour enhancement on measurement point recognition, and it is considered that there are many vertical outlines around the measurement point. Discrete lines using kernel w (k, l) are applied and optimized contrast For reinforcement, real-time local automatic histogram stretching was applied to the results after applying edge reinforcement discrete lines.

1.3. Evaluation results

In order to recognize the measurement points using the developed program, 40 digital tofu measurement radiographs were used, and 40 digital tofu measurement radiographs were divided into system A and system B groups, and system A group was a basic post-process. Through the process, it consists of image-enhanced images. Four calibration physicians identified and input 32 measurement points twice over 40 digital tofu measurement radiographs using the above-described processing techniques 0,1,2,3,4.

Two-factor repeated measurement variance analysis of the error magnitudes of the X and Y coordinates of the processing techniques 0, Y, and Y, which were processed by local image processing technique 0 and local image enhancement technique, respectively. There were measurement points that rejected the null hypothesis and showed significant differences in error magnitudes according to local image processing methods and observers (see FIGS. 9 and 10).

The measurement points showing significant differences were different in the system A group (see FIG. 9) and the system B group (see FIG. 10). Overall, the number of measurement points with significant differences in error magnitude was greater in the System B group. In the case of the system A group, since the image is enhanced through the basic post-processing process in the digital imaging equipment, the difference in effect due to the additional image reinforcement is smaller than that of the group B, and the magnitude of the error is considered to be relatively constant. Even in the system A group, which has already been subjected to the basic image enhancement process, there were some measurement points that showed a significant difference in the magnitude of error caused by the application of the developed image enhancement method. The local image enhancement method based on the image information around the measurement point influences the magnitude of the measurement point recognition error even when the entire image is already enhanced.

The measurement points that showed significant difference in error size according to the image enhancement technique on the X coordinate were basion, PNS, U1A in the system A group, and Pt point, B point, pogonion, gnathion, menton, and soft tissue gnathion in the system B group. According to different results. All the measurement points that showed a significant difference in error magnitude in the system A group showed the smallest horizontal error in the treatment technique 1 (see FIG. 11). In the measurement point of group B, the smallest horizontal error is similar to Pt point and gnathion in treatment technique 3, pogonion and soft tissue gnathion in treatment technique 1, and B point and menton in treatment technique 1 and treatment technique 3 It was shown that (see Figure 12).

The measurement points that showed the significant difference in error size according to the image enhancement technique on the Y coordinate are basion and soft tissue gnathion in the system A group, and A point, pogonion, menton, U1E, L1E, soft tissue A point, soft in the System B group. Tissue gnathion, which also differed slightly from group to group. The measurement point basion of the system A group showed the smallest vertical displacement in the treatment technique 1, and in the case of the soft tissue gnathion, the treatment technique 3 generated more vertical errors (see FIG. 13). A point, L1E for treatment method 2, soft tissue gnathion for treatment method 1, pogonion, menton, U1E, and soft tissue A point for system B group showed the smallest vertical errors. (See Figure 14).

In the measurement point where the magnitude of the error was significantly different, the method showing the smallest error overall was the processing technique 1.

Even if the difference in error reduction is small or statistically insignificant, the application of the real-time local image enhancement technique developed in this study is meaningful. The image around the measurement point is visually superior to the original image and is easier to recognize the measurement point. It converts and expresses in real time, which increases work efficiency.

According to the method for designating a measurement point on the head measurement radiograph according to the present invention, the measurement point can be identified within a locally image-enhanced region.

In addition, according to the method for designating a measurement point on the head measurement radiograph according to the present invention, it is possible to enhance the image locally and in real time.

In addition, according to the method of designating a measurement point on the head measurement radiograph according to the present invention, an image enhancement effect can be enhanced by image enhancement of a digital head measurement radiograph having a DICOM standard format.

Claims (8)

Displaying a digital cephalometric radiograph on a monitor according to a user's request; A second step of image-enhancing a predetermined area including a point where a cursor is positioned on the displayed digital head measurement radiograph; wherein the image enhancement is a real-time localized automatic histogram stretching technique, A second step made by one of a real-time localized automatic histogram equalization technique and an edge enhancement and real-time localized automatic histogram stretching technique; And, And a third step of recognizing a point designated by the user as a measurement point in the image-enhanced constant area. The method of claim 1, wherein the image-enhanced constant region is changed according to the movement of the cursor. 3. A method according to claim 1 or 2, characterized in that the image-enhanced constant area is centered on the cursor. delete delete delete delete 2. The digital towography radiograph of claim 1, wherein the digital towography radiograph before being displayed in the first step has a format that conforms to the DICOM standard, and the image enhancement in the second step is performed on a digital tofugography radiograph that complies with the DICOM standard. Characterized in that the measuring point on the head measurement radiograph.

Images